Color Clustering

# Import Libraries
import os

import pandas as pd
import numpy as np

from sklearn.metrics import confusion_matrix, accuracy_score

from PIL import Image
import cv2
import matplotlib.pyplot as plt
import seaborn as sns

import warnings

# Set Other Configs
warnings.filterwarnings(action="ignore")

pd.options.display.float_format = '{:,.2f}'.format
pd.set_option('display.max_colwidth', None)
pd.set_option('display.max_columns', None)
# pd.set_option('display.max_rows', None)

# Set Path Variables
ROOT_PATH = os.getcwd()
DATA_PATH = os.path.join(ROOT_PATH, "data")

# Create Dataset with Images

# Renkler
colors = ['red', 'blue', 'green', 'gray', 'white']

# Boş bir DataFrame oluştur
df_data = pd.DataFrame(columns=['FileName', 'FilePath', 'Color'])

for color in colors:
    color_path = os.path.join(DATA_PATH, 'roboflow', color)

    # Renk altındaki dosyaları al
    files = os.listdir(color_path)

    # DataFrame'e eklemek için veriyi oluştur
    data = {'FilePath':color_path, 'FileName': files, 'Color': color,}

    # DataFrame'e ekle
    df_data = pd.concat([df_data, pd.DataFrame(data)], ignore_index=True)

df_data['FilePath'] = df_data['FilePath'] + str('\\') + df_data['FileName']

df_data

	FileName	FilePath	Color
0	original1.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\red\original1.jpg	red
1	original10.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\red\original10.jpg	red
2	original11.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\red\original11.jpg	red
3	original12.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\red\original12.jpg	red
4	original13.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\red\original13.jpg	red
...	...	...	...
95	original5.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\white\original5.jpg	white
96	original6.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\white\original6.jpg	white
97	original7.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\white\original7.jpg	white
98	original8.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\white\original8.jpg	white
99	original9.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\white\original9.jpg	white

100 rows × 3 columns

file_names = df_data['FilePath'].to_list()
file_names[0:5]

['C:\\Users\\metin\\Desktop\\YTU\\2-BLM5110-ML\\Homeworks\\blm5110_hw2-clustering\\data\\roboflow\\red\\original1.jpg',
 'C:\\Users\\metin\\Desktop\\YTU\\2-BLM5110-ML\\Homeworks\\blm5110_hw2-clustering\\data\\roboflow\\red\\original10.jpg',
 'C:\\Users\\metin\\Desktop\\YTU\\2-BLM5110-ML\\Homeworks\\blm5110_hw2-clustering\\data\\roboflow\\red\\original11.jpg',
 'C:\\Users\\metin\\Desktop\\YTU\\2-BLM5110-ML\\Homeworks\\blm5110_hw2-clustering\\data\\roboflow\\red\\original12.jpg',
 'C:\\Users\\metin\\Desktop\\YTU\\2-BLM5110-ML\\Homeworks\\blm5110_hw2-clustering\\data\\roboflow\\red\\original13.jpg']

Task 1: Create Image to Array

Description: Resimlerin formatını hazır kütüphaneler kullanarak çözerek görüntüye ait matrisi elde ediniz.Bir görüntü piksellerin (R,G,B) bileşenlerinden oluştuğu bir matristir.

def create_img_to_arr(img_path_list):
    img_rgb_arr = []
    img_bgr_arr = []

    for file in img_path_list:
        img_bgr = cv2.imread(file)
        img_rgb = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2RGB)

        red, green, blue = cv2.split(img_rgb)
        # red, green, blue
        
        height, width, channels = img_rgb.shape
        # height, width, channels

        img_rgb_arr.append(img_rgb)
        img_bgr_arr.append(img_bgr)

    return img_rgb_arr, img_bgr_arr

img_rgb_arr, img_bgr_arr = create_img_to_arr(img_path_list=file_names)

Task 2: Create Color Histogram for each R,G,B Components and Chart

Description: Resimleri oluşturan piksellerin (R,G,B) değerlerine göre her resmin renk histogramını (Her resimde R, G, B bileşenleri için ayrı ayrı olmak üzere toplam 3 histogram) elde ediniz. Histogram eldesi için de hazır kütüphane kullanabilirsiniz. Bir pikselin R,G,B bileşenlerinin değeri 0-255 arası değişti için her histogram dizisi 256 elemanlı olmalıdı.

def img_to_hist_arr(img_arr, is_rgb=True):       
    hist_r_arr = []
    hist_g_arr = []
    hist_b_arr = []
    
    for i in range(0, len(img_arr)):
        img = img_arr[i]
        height, width, channels = img.shape
        
        if is_rgb:
            hist_r = cv2.calcHist([img], [0], None, [256], [0, 256]).ravel()
            hist_g = cv2.calcHist([img], [1], None, [256], [0, 256]).ravel()
            hist_b = cv2.calcHist([img], [2], None, [256], [0, 256]).ravel()
        else:
            hist_r = cv2.calcHist([img], [2], None, [256], [0, 256]).ravel()
            hist_g = cv2.calcHist([img], [1], None, [256], [0, 256]).ravel()
            hist_b = cv2.calcHist([img], [0], None, [256], [0, 256]).ravel()

        hist_r_arr.append(hist_r)
        hist_g_arr.append(hist_g)
        hist_b_arr.append(hist_b)

    return hist_r_arr, hist_g_arr, hist_b_arr

hist_r_arr, hist_g_arr, hist_b_arr = img_to_hist_arr(img_arr=img_rgb_arr, is_rgb=True)

# hist_r_arr, hist_g_arr, hist_b_arr

Plot Histograms RGB Components

def plot_histogram(hist_r_arr, hist_g_arr, hist_b_arr, normalized_msg=""):

    for i in range(0, len(hist_r_arr)):
        hist_r = hist_r_arr[i]
        hist_g = hist_g_arr[i]
        hist_b = hist_b_arr[i]

        plt.figure(figsize=(25, 3))
        plt.subplot(131)
        plt.plot(hist_r, color='red')
        plt.title(f'Red Channel {normalized_msg} Histogram', size=8, loc='left')
        plt.xlim([0, 256])
        
        plt.subplot(132)
        plt.plot(hist_g, color='green')
        plt.title(f'Green Channel {normalized_msg}Histogram', size=8, loc='left')
        plt.xlim([0, 256])
        
        plt.subplot(133)
        plt.plot(hist_b, color='blue')
        plt.title(f'Blue Channel {normalized_msg}Histogram',size=8, loc='left')
        plt.xlim([0, 256])
        
        plt.suptitle(f'Image {i+1} {normalized_msg} Histograms', size=10)
        plt.show()

plot_histogram(hist_r_arr, hist_g_arr, hist_b_arr)

Task 3: Normalizing Color Histogram for each R,G,B Components

Description: Her resimde, her renk bileşeni için histogram dizisindeki sonuçları resimdeki toplam piksel sayısına bölerek dizi elemanlarının değerini [0-1] aralığına normalize ediniz

def img_to_norm_hist_arr(img_arr, is_rgb=True):
    norm_hist_r_arr = []
    norm_hist_g_arr = []
    norm_hist_b_arr = []
    
    for i in range(0, len(img_arr)):
        img = img_arr[i]
        height, width, channels = img.shape
        
        if is_rgb:
            hist_r = cv2.calcHist([img], [0], None, [256], [0, 256]).ravel()
            hist_g = cv2.calcHist([img], [1], None, [256], [0, 256]).ravel()
            hist_b = cv2.calcHist([img], [2], None, [256], [0, 256]).ravel()
        else:
            hist_r = cv2.calcHist([img], [2], None, [256], [0, 256]).ravel()
            hist_g = cv2.calcHist([img], [1], None, [256], [0, 256]).ravel()
            hist_b = cv2.calcHist([img], [0], None, [256], [0, 256]).ravel()
        
        total_pixel = height * width
        norm_hist_r_arr.append(hist_r / total_pixel)
        norm_hist_g_arr.append(hist_g / total_pixel)
        norm_hist_b_arr.append(hist_b / total_pixel)
    
    return norm_hist_r_arr, norm_hist_g_arr, norm_hist_b_arr

norm_hist_r_arr, norm_hist_g_arr, norm_hist_b_arr = img_to_norm_hist_arr(img_arr=img_rgb_arr, is_rgb=True)

Histogram vs Normalized Histogram Chart

plot_histogram(hist_r_arr[0:1], hist_g_arr[0:1], hist_b_arr[0:1], normalized_msg="")

plot_histogram(norm_hist_r_arr[0:1], norm_hist_g_arr[0:1], norm_hist_b_arr[0:1], normalized_msg="Normalized ")

Task 4: K Means Algorithms for k=5

Description: Resimlerin histogramlarını benzerliklerine göre k=5 değeri için(5 renk sınıfı olduğu için) k-means yöntemiyle kümeleyiniz. K-means’de başlangıç adımında rasgele seçeceğiniz k resmin histogramını k cluster’ın başlangıç merkezi olarak kullanınız.

# norm_hist_r_arr, norm_hist_g_arr, norm_hist_b_arr
histograms = np.array([np.concatenate((r, g, b)) for r, g, b in zip(norm_hist_r_arr, norm_hist_g_arr, norm_hist_b_arr)])

KMeans Algorithms

class KMeans:
    def __init__(self, n_clusters=5, max_iters=100, distance_metric='euclidean'):
        self.n_clusters = n_clusters
        self.max_iters = max_iters
        self.distance_metric = distance_metric
        self.centroids = None
        self.clusters = None
        

    def initialize_centroids(self, X):
        centroids_indices = np.random.choice(X.shape[0], self.n_clusters, replace=False)
        centroids = X[centroids_indices]
        return centroids

    def calculate_distance(self, x, y):
        # Uzaklık hesaplama (Öklid veya Manhattan)
        if self.distance_metric == 'euclidean':
            return np.dot(x - y, x - y)
        elif self.distance_metric == 'manhattan':
            return np.sum(np.abs(x - y))
        else:
            raise ValueError("Geçersiz uzaklık metriği. 'euclidean' veya 'manhattan' seçiniz.")

    def fit(self, X):
        self.centroids = self.initialize_centroids(X)

        for _ in range(self.max_iters):
            # Atama aşaması
            self.clusters = np.array([np.argmin([self.calculate_distance(x, y) for y in self.centroids]) for x in X])

            # Güncelleme aşaması
            new_centroids = np.array([X[self.clusters == k].mean(axis=0) for k in range(self.n_clusters)])

            # Eğer merkezler değişmiyorsa döngüyü bitir
            if np.all(self.centroids == new_centroids):
                print("Cluster Converged")
                break

            self.centroids = new_centroids

    def predict(self, X):
        return np.array([np.argmin([self.calculate_distance(x, y) for y in self.centroids]) for x in X])

# # KMeans sınıfını oluştur
kmeans = KMeans(n_clusters=5, max_iters=150, distance_metric='euclidean')

# # Veriyi kümele
kmeans.fit(histograms)

# # Kümeleme sonuçları
clusters = kmeans.predict(histograms)
print(clusters)

Cluster Converged
[1 0 1 0 0 1 1 1 1 1 0 0 1 2 1 1 1 1 4 1 0 0 2 0 0 0 4 0 4 0 0 0 0 0 0 0 0
 0 0 0 0 0 4 0 2 0 2 0 0 0 0 2 0 0 0 2 0 0 0 2 0 4 3 4 0 4 4 0 4 0 4 4 0 0
 4 0 4 0 4 4 0 2 0 1 0 0 1 3 0 0 0 3 0 0 0 0 0 3 0 2]

# for i, cluster in enumerate(clusters):
#     print(f"Resim {i+1} -> Küme {cluster}")

df_data['ClusterId'] = clusters
df_data

	FileName	FilePath	Color	ClusterId
0	original1.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\red\original1.jpg	red	1
1	original10.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\red\original10.jpg	red	0
2	original11.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\red\original11.jpg	red	1
3	original12.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\red\original12.jpg	red	0
4	original13.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\red\original13.jpg	red	0
...	...	...	...	...
95	original5.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\white\original5.jpg	white	0
96	original6.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\white\original6.jpg	white	0
97	original7.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\white\original7.jpg	white	3
98	original8.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\white\original8.jpg	white	0
99	original9.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\white\original9.jpg	white	2

100 rows × 4 columns

Evaluations

Count for each ClusterId

df_data.groupby(by=['ClusterId'], as_index=False).agg(Cnt=('ClusterId', 'count'))

	ClusterId	Cnt
0	0	57
1	1	15
2	2	9
3	3	4
4	4	15

Count for each ClusterId and Color

df_data.groupby(by=['ClusterId', 'Color'], as_index=False).agg(Cnt=('ClusterId', 'count')).sort_values(by=['Cnt'], ascending=False)

	ClusterId	Color	Cnt
0	0	blue	17
2	0	green	14
4	0	white	13
5	1	red	13
14	4	gray	11
1	0	gray	8
8	2	green	5
3	0	red	5
12	3	white	3
6	1	white	2
10	2	white	2
13	4	blue	2
7	2	blue	1
9	2	red	1
11	3	gray	1
15	4	green	1
16	4	red	1

Cluster Performance Evaluation

def cluster_evaluate(df, label_1="", label_2="", labels=[], visualise=False):
    accuracy = accuracy_score(df[label_1], df[label_2])
    print(f"Accuracy: {accuracy}")
    if visualise:
        conf_matrix = confusion_matrix(df[label_1], df[label_2], labels=labels)
        plt.figure(figsize=(8, 6))
        sns.heatmap(conf_matrix, annot=True, fmt='d', cmap='Blues', xticklabels=['red', 'blue', 'green', 'white', 'gray'], yticklabels=['red', 'blue', 'green', 'white', 'gray'])
        plt.xlabel('Tahmin Edilen Renk')
        plt.ylabel('Gerçek Renk')
        plt.title('Confusion Matrix')
        plt.show()

Visualize for each ClusterId

import matplotlib.pyplot as plt
import matplotlib.image as mpimg
import numpy as np

def display_images(image_paths, images_per_row):
    num_images = len(image_paths)
    num_rows = int(np.ceil(num_images / images_per_row))

    plt.figure(figsize=(15, 3 * num_rows))  # Figür boyutunu ayarlama
    for i, image_path in enumerate(image_paths):
        # img = mpimg.imread(image_path)  # Görseli yükleme
        img = cv2.imread(image_path)  # Görseli yükleme
        img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
        plt.subplot(num_rows, images_per_row, i + 1)  # Satır satır ve belirlenen sayıda sütunda i+1. görseli ayarlama
        plt.imshow(img)
        f = str(image_path.split("\\")[-2:][0]) + str("-") + str(image_path.split("\\")[-2:][1])
        plt.title(f'{f}', size=8, loc='center')
        plt.axis('off')  # Eksenleri kapatma

    plt.tight_layout()  # Layout'u düzenleme
    plt.show()

cluster_color_dict = df_evaluate[['ClusterId', 'ClusterColor']].drop_duplicates().sort_values(by=['ClusterId'], ascending=True).reset_index(drop=True).to_dict()['ClusterColor']    
for i in range(0, 5):
    color_name = cluster_color_dict[i]
    print("Cluster Id:{} & Color:{}".format(i, color_name))
    # image_paths = ['resim{}.jpg'.format(i) for i in range(1, 101)]  # örnek isimler
    image_paths = df_data[df_data['ClusterId']==i]['FilePath'].to_list()
    
    X = 10  # Her satırdaki resim sayısı
    display_images(image_paths, X)
    print("= " * 25)

Cluster Id:0 & Color:blue

= = = = = = = = = = = = = = = = = = = = = = = = = 
Cluster Id:1 & Color:red

= = = = = = = = = = = = = = = = = = = = = = = = = 
Cluster Id:2 & Color:green

= = = = = = = = = = = = = = = = = = = = = = = = = 
Cluster Id:3 & Color:white

= = = = = = = = = = = = = = = = = = = = = = = = = 
Cluster Id:4 & Color:gray

= = = = = = = = = = = = = = = = = = = = = = = = = 

Task : Deneysel Sonuçlar

10 Tekrar ile Kümeleme Performansının Ölçülmesi

a. Küme merkezlerini rasgele belirleyerek her k değeri için kümeleme işlemini 10 defa tekrarlayınız. Her kümeleme sonunda, kümelerde doğru cluster’da olan resim yüzdesini hesaplayınız.

# df_cluster_result = pd.DataFrame()
df_cluster_result = df_data[['FileName', 'FilePath','Color']]
for i in range(0, 10):
    kmeans = KMeans(n_clusters=5, max_iters=150)
    kmeans.fit(histograms)
    clusters = kmeans.predict(histograms)
    df_cluster_result[f'ClusterId_{i+1}'] = clusters

df_cluster_result

Cluster Converged
Cluster Converged
Cluster Converged
Cluster Converged
Cluster Converged
Cluster Converged
Cluster Converged
Cluster Converged
Cluster Converged
Cluster Converged

	FileName	FilePath	Color	ClusterId_1	ClusterId_2	ClusterId_3	ClusterId_4	ClusterId_5	ClusterId_6	ClusterId_7	ClusterId_8	ClusterId_9	ClusterId_10
0	original1.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\red\original1.jpg	red	1	4	4	0	4	0	0	2	4	0
1	original10.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\red\original10.jpg	red	1	4	4	0	4	0	0	2	4	0
2	original11.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\red\original11.jpg	red	1	4	4	0	4	0	0	2	4	0
3	original12.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\red\original12.jpg	red	1	4	2	0	4	4	0	2	4	0
4	original13.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\red\original13.jpg	red	1	4	4	0	4	4	0	2	4	0
...	...	...	...	...	...	...	...	...	...	...	...	...	...
95	original5.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\white\original5.jpg	white	1	4	2	0	4	4	0	2	4	0
96	original6.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\white\original6.jpg	white	1	4	2	0	4	4	0	2	4	0
97	original7.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\white\original7.jpg	white	3	3	1	1	2	3	3	3	1	3
98	original8.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\white\original8.jpg	white	1	4	2	0	4	4	0	2	4	0
99	original9.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\white\original9.jpg	white	0	0	3	2	3	1	2	0	2	4

100 rows × 13 columns

for i in range(0, 10):
    df_multi_cluster = df_cluster_result.groupby(by=[f'ClusterId_{i+1}'], as_index=False)['Color'].apply(lambda x: x.value_counts().idxmax())
    df_multi_cluster.rename(columns={'Color':f'ClusterColor_{i+1}'}, inplace=True)
    df_cluster_result[f'ClusterColor_{i+1}'] = df_cluster_result[f'ClusterId_{i+1}'].map(df_multi_cluster.to_dict()[f'ClusterColor_{i+1}'])
    print(f"Cluster_{i+1}")
    cluster_evaluate(df=df_cluster_result, label_1="Color", label_2=f"ClusterColor_{i+1}", labels=['red', 'blue', 'green', 'white', 'gray'], visualise=False)
    print("* " * 10)

Cluster_1
Accuracy: 0.54
* * * * * * * * * * 
Cluster_2
Accuracy: 0.38
* * * * * * * * * * 
Cluster_3
Accuracy: 0.5
* * * * * * * * * * 
Cluster_4
Accuracy: 0.37
* * * * * * * * * * 
Cluster_5
Accuracy: 0.33
* * * * * * * * * * 
Cluster_6
Accuracy: 0.49
* * * * * * * * * * 
Cluster_7
Accuracy: 0.38
* * * * * * * * * * 
Cluster_8
Accuracy: 0.32
* * * * * * * * * * 
Cluster_9
Accuracy: 0.38
* * * * * * * * * * 
Cluster_10
Accuracy: 0.33
* * * * * * * * * * 

for i in range(0, 10):
    df_multi_cluster = df_cluster_result.groupby(by=[f'ClusterId_{i+1}'], as_index=False)['Color'].apply(lambda x: x.value_counts().idxmax())
    df_multi_cluster.rename(columns={'Color':f'ClusterColor_{i+1}'}, inplace=True)
    df_cluster_result[f'ClusterColor_{i+1}'] = df_cluster_result[f'ClusterId_{i+1}'].map(df_multi_cluster.to_dict()[f'ClusterColor_{i+1}'])
    print(f"Cluster_{i+1}")
    cluster_evaluate(df=df_cluster_result, label_1="Color", label_2=f"ClusterColor_{i+1}", labels=['red', 'blue', 'green', 'white', 'gray'], visualise=True)
    print("* " * 10)

Cluster_1
Accuracy: 0.54

* * * * * * * * * * 
Cluster_2
Accuracy: 0.38

* * * * * * * * * * 
Cluster_3
Accuracy: 0.5

* * * * * * * * * * 
Cluster_4
Accuracy: 0.37

* * * * * * * * * * 
Cluster_5
Accuracy: 0.33

* * * * * * * * * * 
Cluster_6
Accuracy: 0.49

* * * * * * * * * * 
Cluster_7
Accuracy: 0.38

* * * * * * * * * * 
Cluster_8
Accuracy: 0.32

* * * * * * * * * * 
Cluster_9
Accuracy: 0.38

* * * * * * * * * * 
Cluster_10
Accuracy: 0.33

* * * * * * * * * * 

Cluster Performansının Ölçülmesi

b. Hazırlayacağınız karışıklık matrisi (confusion matrix) üzerinde sonucu gösteriniz

df_cluster = df_data.groupby(by=['ClusterId'], as_index=False)['Color'].apply(lambda x: x.value_counts().idxmax())
df_cluster.rename(columns={'Color':'ClusterColor'}, inplace=True)
df_cluster

	ClusterId	ClusterColor
0	0	blue
1	1	red
2	2	green
3	3	white
4	4	gray

df_evaluate = pd.merge(df_data, df_cluster, on='ClusterId', how='left')
df_evaluate

	FileName	FilePath	Color	ClusterId	ClusterColor
0	original1.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\red\original1.jpg	red	1	red
1	original10.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\red\original10.jpg	red	0	blue
2	original11.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\red\original11.jpg	red	1	red
3	original12.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\red\original12.jpg	red	0	blue
4	original13.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\red\original13.jpg	red	0	blue
...	...	...	...	...	...
95	original5.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\white\original5.jpg	white	0	blue
96	original6.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\white\original6.jpg	white	0	blue
97	original7.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\white\original7.jpg	white	3	white
98	original8.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\white\original8.jpg	white	0	blue
99	original9.jpg	C:\Users\metin\Desktop\YTU\2-BLM5110-ML\Homeworks\blm5110_hw2-clustering\data\roboflow\white\original9.jpg	white	2	green

100 rows × 5 columns

cluster_evaluate(df=df_evaluate, label_1="Color", label_2="ClusterColor", labels=['red', 'blue', 'green', 'white', 'gray'], visualise=True)

Accuracy: 0.49

Doğru ve Yanlış Clusterların Görselleştirilmesi

c. Her sınıf için doğru cluster’a yerleşmiş 5 örnek ve yanlış cluster’da bulunan 1 örnek resmi veriniz.

# Doğru Yerleşmiş Örnekler
cluster_color_dict = df_evaluate[['ClusterId', 'ClusterColor']].drop_duplicates().sort_values(by=['ClusterId'], ascending=True).reset_index(drop=True).to_dict()['ClusterColor']
for i in range(0, 5):
    color_name = cluster_color_dict[i]
    print("Cluster Id:{} & Color:{}".format(i, color_name))
    image_paths = df_evaluate[(df_evaluate['ClusterId']==i) & (df_evaluate['Color']==df_evaluate['ClusterColor'])]['FilePath'].head(5).to_list()
    
    X = 10  # Her satırdaki resim sayısı
    display_images(image_paths, X)
    print("= " * 25)

Cluster Id:0 & Color:blue

= = = = = = = = = = = = = = = = = = = = = = = = = 
Cluster Id:1 & Color:red

= = = = = = = = = = = = = = = = = = = = = = = = = 
Cluster Id:2 & Color:green

= = = = = = = = = = = = = = = = = = = = = = = = = 
Cluster Id:3 & Color:white

= = = = = = = = = = = = = = = = = = = = = = = = = 
Cluster Id:4 & Color:gray

= = = = = = = = = = = = = = = = = = = = = = = = = 

# Yanlış Yerleşmiş Örnekler
cluster_color_dict = df_evaluate[['ClusterId', 'ClusterColor']].drop_duplicates().sort_values(by=['ClusterId'], ascending=True).reset_index(drop=True).to_dict()['ClusterColor']
for i in range(0, 5):
    color_name = cluster_color_dict[i]
    print("Cluster Id:{} & Color:{}".format(i, color_name))
    image_paths = df_evaluate[(df_evaluate['ClusterId']==i) & (df_evaluate['Color']!=df_evaluate['ClusterColor'])]['FilePath'].head(5).to_list()
    
    X = 10  # Her satırdaki resim sayısı
    display_images(image_paths, X)
    print("= " * 25)

Cluster Id:0 & Color:blue

= = = = = = = = = = = = = = = = = = = = = = = = = 
Cluster Id:1 & Color:red

= = = = = = = = = = = = = = = = = = = = = = = = = 
Cluster Id:2 & Color:green

= = = = = = = = = = = = = = = = = = = = = = = = = 
Cluster Id:3 & Color:white

= = = = = = = = = = = = = = = = = = = = = = = = = 
Cluster Id:4 & Color:gray

= = = = = = = = = = = = = = = = = = = = = = = = =